skill-creator: skills from OpenAPI, GraphQL and MCP
@asnd/skill-creator
The /skill-creator command turns a link to an OpenAPI spec, GraphQL schema or MCP server into a ready skill with wrappers.
- 622 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- A generated wrapper calls the API with your tokens from environment variables and can send requests that change service data
- For MCP over stdio the server start command is passed as a string, and wrappers run the package through npx on every call
- The generator writes files into agent directories and keeps a `lock.json` file in your home folder, and the improvement skill may edit skills listed there
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/asnd-skill-creatorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- A generated wrapper calls the API with your tokens from environment variables and can send requests that change service data
- For MCP over stdio the server start command is passed as a string, and wrappers run the package through npx on every call
- The generator writes files into agent directories and keeps a `lock.json` file in your home folder, and the improvement skill may edit skills listed there
This is third-party code. Review the repository files before installing.
What it does
The package installs a `/skill-creator` command into your agent. You give it a link to an OpenAPI spec, a GraphQL schema or an MCP server, and the agent studies the source, runs the generator and assembles a ready skill. The skill holds a SKILL.md file, a wrapper script and a copy of the spec in a `references` folder. Later sessions do not have to reread the API docs: they call the wrapper. The wrapper has four modes: `commands list` shows the commands, `commands search` searches by a word, `commands help` explains one command and `run` executes it. Secrets come from environment variables or files and never reach the generated documents. Specs, schemas and MCP tool lists are cached, and flags set the lifetime and force a refresh. The command ships with the skill-creator-improvement skill. It appends to the Gotchas section of a generated skill whatever the agent learned in use: custom fields, corrected commands, faster call chains. It may change only skills recorded in the `~/.skill-creator/lock.json` file.
Who it is for. For developers who work with a coding agent and want to package access to their API or MCP server as a skill once.
Good fit when
- An internal service has an OpenAPI spec and the agent should call it without pasting docs into every chat
- You want to turn an MCP server into a skill with commands, so its tools are not kept in context all the time
- You have a GraphQL endpoint and the agent needs ready commands for queries with field selection
- A team shares one skill across several agents: installation covers Claude Code, Codex, Cursor and others
- A service skill should get better with use, with fixes accumulating in the Gotchas section
Not a fit when
- You need a persistent MCP server with subscriptions and notifications: a generated skill calls commands on request
- The service has no OpenAPI, GraphQL or MCP: the generator would have no source
- You need a skill with an author's own working method: the generator builds a wrapper around an API, and you must write the method yourself
- An environment without a suitable Node.js: the package requires Node.js ^22.22.2, ^24.15.0 or 26 and newer
Example request
/skill-creator --spec https://example.com/openapi.json --name youtube --agent claude-code --scope projectLimitations
It needs Node.js ^22.22.2, ^24.15.0 or 26 and newer. Scripts are generated for the chosen agent: pi, claude-code, codex, cursor, opencode, gemini-cli, github-copilot, cline and windsurf are supported, and the `--scope` flag sets the install scope to global or project. Generated wrappers run the package through `npx -y @asnd/skill-creator`, so every call needs npm access or a warm cache.
The skill's quality depends on the source: a spec with poor descriptions yields awkward commands, and SKILL.md will need polishing. For an MCP source over stdio, the server start command is passed as a string and runs on your computer. The package version is 0.2.0 and the repository was last updated in June 2026.
Whether npm and the services you connect the generator to are reachable from Russia was not checked for this card.
How to disable. Delete the command file and the skill-creator-improvement skill folder from the agent's directory, for example ~/.pi/agent/prompts and ~/.pi/agent/skills for pi. Generated skills live in their own folders and are removed together with their entries in `~/.skill-creator/lock.json`. The cache is kept in `~/.cache/skill-creator`.
FAQ
How do I install /skill-creator in Claude Code?
Run `npx @asnd/skill-creator command install --agent claude-code --scope project` and the command appears in the project. The `--scope global` flag installs it for all projects, and `--no-improvement-skill` skips the companion improvement skill.
Which sources does the generator accept?
An OpenAPI spec by file or URL, a GraphQL endpoint with an optional SDL schema or introspection result, an MCP server over streamable HTTP or SSE, and an MCP server over stdio started by a command string. The templates are called openapi, graphql, mcp-http and mcp-stdio.
What is in a generated skill?
A `SKILL.md` file, a `scripts` folder with the wrapper script and a `references` folder with a copy of the spec, whose file name carries a date. The wrapper is called as a script with the subcommands `commands list`, `commands search`, `commands help` and `run`.
How do I pass an API token so it does not end up in the skill?
With the `--auth-header` flag: `Header:env:NAME` reads the secret from an environment variable and `Header:file:/path/token` reads it from a file. The secret stays out of the generated documents, and the flag is repeatable.
What does the skill-creator-improvement skill do?
After you use a generated skill, it appends verified findings to the Gotchas section: correct commands, custom fields, pagination patterns. It checks the `lock.json` file and leaves untracked skills alone.
Discussion
No questions or reviews yet.
Related
CLI toolsBuilder Skills
Skills for agents
A set of small skills from Builder.io for agent work habits: visual plans and diff recaps, docs-first checks, cross-agent review and clean status updates
The author's personal set: shared agent rules and dozens of skills for GitHub, the browser, macOS, messaging and releases, plus a sync script
Tech Leads Club Agent Skills
Agent Skills
A vetted skill registry with a CLI that installs into twenty-plus agents and an MCP server for on-demand skill lookup
Dimillian Skills
Skills
A set of 16 engineering skills for agents: Apple platform work, macOS packaging, React performance, code review and refactoring
Softaworks Agent Toolkit
Softaworks Agent Skills
A marketplace of skills, subagents and commands for coding agents: development, documentation, planning and workplace communication
OpenClaw Agent Skills
openclaw-agent-skills
Shared skills for coding agents that work on OpenClaw projects: review and other workflows. A narrow topic for people who write OpenClaw code
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| skill-creator: skills from OpenAPI, GraphQL and MCP (this entry) | no | 622 | |
| mcp2cli | no | 2,425 | turns MCP, OpenAPI or GraphQL into commands on the fly, without packaging a separate skill |
| OpenAPI to CLI (ocli) | no | 267 | makes CLI commands from OpenAPI and Swagger without code generation |
README in short
The document explains the idea: instead of pasting API docs into every chat, the agent gets a skill with ready commands. It describes installing the command for various agents, launch examples for OpenAPI, GraphQL and MCP, the structure of a generated skill and the improvement loop with the `lock.json` file. A flag table covers authorization, command filters, MCP transport choice and caching. For CI it gives direct generate calls with templates and the list of pnpm development commands.